Spline-based resampling of noisy images
Atanas P. Gotchev, Karen Egiazarian · 2002
We consider the problem of image resampling in the presence of noise in terms of a regularized solution for spline-like model coefficients. The properties of the generalized cross-validation (GCV) and the Akaike information criterion (AIC) for determination of the time number of the model coefficients and the value of the regularization parameter have been examined. A two-parameter optimization procedure can be applicable in the case of noisy resampling. The method is applicable when the image should be resampled and denoised at the same time. The problem is somehow related to the problem of finding the natural scale of representation, when one aims to find the image scale optimum for further processing (compression, denoising, etc.). The practical realization of the method is discussed as well.